AEO Page Rank: How Individual Pages Are Scored for Citation Readiness

Here is a scenario we see constantly. A site scores 72 on AEO Site Rank. Decent infrastructure. Good schema. Solid llms.txt file. The site owner assumes their content is in good shape.

Then we score their individual pages. The homepage? 78. The FAQ page? 81. The blog post they spent two weeks writing? 39.

That blog post - the one they are counting on to drive AI citations for their core topic - has no original data, no named entities beyond generic industry terms, sentences averaging 35 words each, and every other paragraph starts with "As we discussed above." AI engines cannot extract a single clean answer from it.

AEO Site Rank and AEO Page Rank exist because they answer different questions:

AEO Site Rank asks: "Is this site well-built for AI visibility?" It evaluates 48 criteria across your entire domain - schema markup, crawlability, llms.txt, robots.txt, topic coherence, content depth patterns, discovery signals. These are infrastructure-level concerns. They determine whether AI engines can find and trust your site.

AEO Page Rank asks: "Is this specific page ready to be cited?" It evaluates 17 checks on a single page - whether the content incorporates the gathered corpus, whether it says something original, whether answers can be cleanly extracted, whether entities and data are properly referenced, and whether the structure helps machine comprehension.

A site with excellent infrastructure but weak individual pages will have a high AEO Site Rank and low AEO Page Ranks. AI engines can find the site. They just cannot find anything worth citing on specific pages.

Flip it. A brilliant, data-rich blog post sitting on a site with no schema, blocked AI crawlers, and no llms.txt will have a high AEO Page Rank and a low AEO Site Rank. The content is citation-ready. The site makes it nearly impossible for engines to discover it.

You need both numbers because the fix is different in each case. Infrastructure problems require technical work. Content problems require editorial work. Conflating the two wastes time and money.

AEO Page Rank scores a single page from 0 to 100 using 17 deterministic checks grouped into 5 pillars. No LLM opinions. No prompt sensitivity. Run it twice on the same page and you get the same number.

The five pillars, in order of weight:

Content Originality (25%) - One check measures whether the finished page actually uses the intelligence gathered upstream: - Content originality - Cross-references the article against the research corpus and synthesized artifact. If the page barely reflects the collected source material, it loses a full quarter of the score.

Content Uniqueness (25%) - Four checks measure whether the page contains information AI engines cannot find elsewhere: - Owned insight density - Does the page contain "our analysis," "our data," "our testing" patterns? This is provenance-aware: if you supply evidence metadata proving your claims, the score scales higher. Without provenance, it caps at 50%. - Named framework - Does the page introduce a branded methodology? "The CLEAR Framework" or "Our 5-Step Protocol" signals original intellectual property. - Contrarian claims - Does the page challenge conventional wisdom? "Unlike what most guides suggest" or "This contradicts the standard approach" patterns indicate original thinking. - Novel synthesis - Does the page combine data from multiple sources into something new? Cross-source comparison tables and aggregation signals score here.

Extractability (25%) - Can AI engines pull clean answers from this page? - Answer capsule - After each H2 heading, is there a self-contained 15-30 word answer? No links, no dangling references. Just a clean statement an engine can lift verbatim. - Information island - What percentage of paragraphs are self-contained? Paragraphs starting with "This," "That," or "As mentioned above" are not extractable because they depend on surrounding context. - Atomic sentences - What percentage of sentences make a single claim in under 20 words? Long compound sentences with multiple claims are harder for engines to cite accurately. - Front-loading - Is the signal density concentrated in the first 30% of the page? AI engines often extract from early content. A page that buries its best insights at the bottom scores lower.

Entity & Data Richness (15%) - Does the page reference real things with real numbers? - Entity density - Proper noun ratio. Cited text averages 20.6% entity density. If your page is all generic nouns without naming specific companies, people, tools, or places, engines have nothing concrete to anchor citations to. - Fact density - Numeric data points with context. "Revenue grew 34% year-over-year" scores. "Revenue grew significantly" does not. - Evidence packaging - "According to X" attribution phrases. Explicit source attribution makes content more trustworthy and citable. - Freshness signals - Content dates, update timestamps, and visible date markers. Stale content without dates gets deprioritized.

Structural Signals (10%) - Is the page formatted for machine comprehension? - Question H2 ratio - What percentage of H2 headings are phrased as questions? Target is 70%. Question headings directly match how users query AI engines. - Definition density - "X is..." patterns. Clear definitions are among the most-cited content patterns. - Schema completeness - JSON-LD structured data on the page. Article schema, FAQ schema, breadcrumb schema. - Subjectivity balance - The ratio of factual to interpretive content. Advisory only (lower confidence) - pure fact pages and pure opinion pages both score lower than pages that blend data with analysis.

The two scores share a name pattern but almost nothing else. Different scope. Different criteria. Different pillar structure. Different use cases.

                        AEO Site Rank              AEO Page Rank
-----------------------------------------------------------------------
Scope                   Entire domain              Single page
Criteria count          48                         17
Pillars                 5                          5
Heaviest pillar         Answer Readiness           Originality / Uniqueness / Extractability
Key question            "Is the site built          "Is this page ready
                         for AI visibility?"         to be cited?"
Infrastructure focus    High (robots.txt,          None (page content
                         llms.txt, schema,           only)
                         crawlability)
Content focus           Moderate (coherence,       High (corpus usage,
                         depth patterns)             originality, extractability)
Page sampling           Samples fleet of pages     Scores one page
Coherence gate          Yes (caps site score)      No (not applicable)
Duplicate content       Part of scoring            Veto gate (caps score)
Score range             0-100                      0-100
Grade scale             No letter grades           A/B/C/D/F grades

Here is the key insight: 12 of AEO Page Rank's 17 checks do not exist in AEO Site Rank at all. Content originality, owned insight density, named frameworks, contrarian claims, novel synthesis, answer capsules, information islands, atomic sentences, front-loading, entity density, question H2 ratio, and subjectivity balance are all unique to AEO Page Rank. They measure citation readiness at a granularity that site-level scoring cannot reach.

The remaining 5 checks overlap conceptually with AEO Site Rank. Under the hood they reuse six scorer functions: schema completeness, content freshness, visible date signals, definition patterns, fact density, and evidence packaging. These are criteria that matter at both levels - a site needs good schema patterns and a page needs schema on it specifically.

The duplicate content gate works differently too. In AEO Site Rank, duplicate content is one of 48 weighted criteria. In AEO Page Rank, it is a veto gate. If a page scores 6 or below on duplicate content, the entire AEO Page Rank is capped regardless of how well the 17 weighted checks perform:

Duplicate Content Score    AEO Page Rank Cap
--------------------------------------------
0                          35 maximum
1-2                        40-45 maximum
3-4                        50-55 maximum
5-6                        60-65 maximum
7+                         No cap

This is deliberate. A page that recycles content from other sources is not citation-ready no matter how well-structured it is. AI engines already have the original. They do not need your copy of it.

AEO Page Rank uses letter grades alongside the 0-100 score. The grade tells you at a glance whether a page is ready for AI citation or needs work.

Grade A (85-100): Citation-ready. This page contains original insights, clean extractable answers, rich entity references, and proper structural formatting. AI engines can confidently cite specific statements from this page. Pages in this range typically have named frameworks or proprietary data, self-contained answer capsules after every major heading, high entity density, and question-formatted H2s.

Grade B (70-84): Strong with minor gaps. The page is structurally sound and contains citable content. One or two pillars might trail - maybe the extractability is strong but the content uniqueness could be sharper. A Grade B page will get cited in some contexts but may lose out to a Grade A competitor page on the same topic.

Grade C (55-69): Moderate - needs content work. The page has some citable elements but significant gaps. Common patterns: decent structure but generic content (no original data), or original insights buried in long paragraphs that AI cannot cleanly extract. Grade C pages get cited occasionally but inconsistently.

Grade D (40-54): Weak citation readiness. AI engines can read this page but struggle to find anything worth citing. The content is either too generic, too poorly structured, or both. Most marketing landing pages and thin product pages land here. The fix usually requires both new content (original data, named entities) and structural improvements (answer capsules, atomic sentences).

Grade F (0-39): Not citation-ready. This page is effectively invisible to AI citation. It may exist in search indexes, but when an engine needs to cite a source for a topic this page covers, it will choose a competitor every time. Common causes: duplicated content (triggers the veto gate), zero original data, no extractable answer patterns, and missing structural signals.

The passing threshold is 70 - Grade B or above. That does not mean pages below 70 are worthless. It means they are not reliably citable. For pages that matter to your business - your core topic pages, your pillar content, your conversion pages - you want Grade B or higher.

The Studio dashboard shows both scores side by side for a reason. They create a two-dimensional map of your AI visibility position.

Quadrant 1: High Site Rank + High Page Ranks. This is the goal state. Your infrastructure makes you discoverable, and your content makes you citable. AI engines can find your site and find things worth quoting on it. Keep publishing and monitoring.

Quadrant 2: High Site Rank + Low Page Ranks. Your site is technically excellent - great schema, proper crawlability, comprehensive llms.txt. But the actual content on individual pages is generic, poorly structured, or thin. This is the "beautiful empty house" problem. AI engines visit. They just leave without citing anything. Fix: editorial work on your highest-traffic pages. Add original data, restructure for extractability, increase entity density.

Quadrant 3: Low Site Rank + High Page Ranks. You have brilliant content that nobody can find. Individual pages are packed with original insights, clean answer capsules, and rich data. But the site infrastructure - missing schema, blocked crawlers, no llms.txt, weak internal linking - prevents AI engines from discovering those pages. Fix: technical AEO work. The content is ready. The plumbing is not.

Quadrant 4: Low Site Rank + Low Page Ranks. Both infrastructure and content need work. Start with the site-level fixes first (they affect every page simultaneously) then work on individual pages. Adding llms.txt, fixing robots.txt, and implementing basic schema will lift the site score. Then focus content improvements on your 3-5 most important pages.

In the Studio, each content card shows the AEO Page Rank as a score badge with a "PR" label to distinguish it from the site-level AEO Rank. The preview panel runs a live analysis that returns both scores, so you can see exactly how a page edit affects citation readiness without waiting for a full site re-audit.

The content pipeline also uses AEO Page Rank at three checkpoints during article generation: before humanization (quality gate), after humanization (preservation check), and on the final published HTML (definitive score). If a generated article drops below 70 at any checkpoint, the pipeline flags it for review.

The weight distribution tells you where to focus. Content Originality, Content Uniqueness, and Extractability are each 25% of the score. That means 75% of your AEO Page Rank comes from using the research corpus well, saying something distinctive, and making it easy for AI engines to extract. Start there.

Use the corpus aggressively (Content Originality - 25%). A quarter of the score now depends on whether the finished page actually reflects the intelligence you gathered upstream. If research surfaced statistics, expert quotes, competitor patterns, or synthesized themes, they need to show up in the article. A generic page that ignores its source bundle leaves too many points on the table.

Add original data and interpretation (Content Uniqueness - 25%). The next major lever. Stop citing the same industry statistics everyone else uses without adding anything new. AI engines already have those. Instead, publish your own data: customer survey results, internal benchmarks, A/B test outcomes, operational metrics. "Our analysis of 1,200 customer support tickets found that..." is original. "According to a recent study, 73% of customers prefer..." is not.

Create named frameworks. If you have a methodology, name it. "The 4-Step Content Audit Protocol" or "Our RISE Framework for AI Visibility." Named frameworks signal intellectual property that AI engines cannot attribute to anyone else.

Write answer capsules after every H2 (Extractability - 25%). Immediately after each H2 heading, write a 15-30 word self-contained statement that answers the heading's question. No links. No "this" or "that" references. A clean sentence an AI engine can lift verbatim.

Make paragraphs self-contained. Every paragraph should make sense if read in isolation. Delete "As mentioned above," "This approach," "That said." Replace them with the actual thing being referenced. AI engines extract individual paragraphs. If a paragraph depends on the one before it, it is invisible to extraction.

Shorten sentences. Target single-claim sentences under 20 words. "We reduced customer churn by 12% in Q3 2025" is atomic and citable. "By implementing a comprehensive customer success program that included automated health scoring, proactive outreach, and quarterly business reviews, we managed to significantly reduce our overall customer churn rates" is not.

Front-load your best content. Put your strongest claims, most original data, and clearest answers in the first 30% of the page. AI engines disproportionately extract from early content. A page that opens with three paragraphs of background context before getting to the point scores lower on front-loading.

Increase entity density (Entity & Data Richness - 20%). Name specific companies, people, tools, frameworks, and locations. "Leading CRM platforms" is generic. "Salesforce, HubSpot, and Pipedrive" is entity-rich. Cited text averages 20.6% entity density. If your page is below 15%, you are underperforming.

Use question-formatted H2s (Structural Signals - 10%). "How Does Customer Churn Affect Revenue?" is better than "The Impact of Customer Churn." Question H2s directly match how users query AI engines. Target 70% of your H2s as questions.

Start with your three most important pages. Score them with AEO Page Rank. Fix the lowest-scoring pillar on each. Re-score. The feedback loop is immediate - every change maps directly to a specific check.

• AEO Site Rank Methodology - www.aeocontent.ai/knowledge/aeo-score-methodology
• AEO Site Rank: Benchmarks and Peer Comparison - www.aeocontent.ai/knowledge/aeo-rank-methodology
• AEO Page Rank API - www.aeocontent.ai/developers
• AEO Rankings - All Sectors - www.aeocontent.ai/rankings